Advanced Compiler Design. CSE 231 Instructor: Sorin Lerner

Transcription

1 Advanced Compiler Design CSE 231 Instructor: Sorin Lerner

2 Let s look at a compiler if ( ) { x := ; } else { y := ; } ; Parser Compiler Compiler Optimizer Code Gen Exec

3 Let s look at a compiler Compiler Parser Optimizer Code Gen

4 Advanced Optimizer Design CSE 231 Instructor: Sorin Lerner

5 What does an optimizer do? Compiler Parser Optimizer Code Gen 1. Compute information about a program 2. Use that information to perform program transformations (with the goal of improving some metric, e.g. performance)

6 What do these tools have in common? Bug finders Program verifiers Code refactoring tools Garbage collectors Runtime monitoring system And optimizers

7 What do these tools have in common? Bug finders Program verifiers Code refactoring tools Garbage collectors Runtime monitoring system And optimizers They all analyze and transform programs We will learn about the techniques underlying all these tools

8 Program Analyses, Transformations, and Applications CSE 231 Instructor: Sorin Lerner

9 Course goals Understand basic techniques cornerstone of a variety of program analysis tools useful no matter what your future path Get a feel for compiler research/implementation useful for research-oriented students useful for implementation-oriented students

10 Course topics Representing programs Analyzing and transforming programs Applications of these techniques

11 Course topics (more details) Representations Abstract Syntax Tree Control Flow Graph Dataflow Graph Static Single Assignment Control Dependence Graph Program Dependence Graph Call Graph

12 Course topics (more details) Analysis/Transformation Algorithms Dataflow Analysis Interprocedural analysis Pointer analysis Rule-based analyses and transformations Constraint-based analysis

13 Course topics (more details) Applications Scalar optimizations Loop optimizations Object oriented optimizations Program verification Bug finding

14 Course pre-requisites No compilers background necessary No familiarity with lattices I will review what is necessary in class Familiarity with functional/oo programming Optimization techniques for these kinds of languages Familiarity with C/C++ Project will be in C++ Standard ugrad cs curriculum likely enough Talk to me if you re concerned

15 Course work In-class midterm (30%) Date posted on web site In-class final (30%) Date published by official calendar Course project (35%) Class participation (5%)

16 Course project Goal of the project Get some hands on experience with compilers Two options, most will do option 1 Option 1: LLVM project Implement some analyses in LLVM, three milestones Hand in your code and it s auto-graded Option 2: Research (by instructor approval) Pick some interesting idea, and try it out Proposals due at the beginning of the second week Can leverage your existing research

17 LLVM Project M1: Simple instrumentation M2: Analysis framework M3: Implement Analyses in framework You will extend LLVM. This will require C++ If you don t know C++, you will learn If you don t think you can learn C++, think about dropping this class? To be done alone

18 Research Project Requires instructor approval You need to come up with your own idea by the end of week 1 Most students doing this will be PhD students It s ok to leverage or overlap with existing research I envision at most 10 people doing this

19 Readings Paper readings throughout the quarter Seminal papers and state of the art Gives you historical perspective Shows you lineage from idea to practice

20 Administrative info Class web page is up (or Google Sorin Lerner, follow Teaching Now ) Will post lectures, readings, project info, etc. Piazza link on web page Use for questions, answers Especially LLVM/project Q&A

21 Academic Integrity Governed by Policy on Integrity of Scholarship ( Manual/Appendices/2) Allegations are handled by Academic Integrity Office ( Course penalty for any cheating in 231 will be a failing grade for the entire class Cheaters may be subject to additional administrative sanctions

22 Questions?

23 Program Analyzer Issues (discuss) Input Program Analyzer Output

24 Program Analyzer Issues (discuss) Input Program Analyzer Output

25 Program Analyzer Issues (discuss) Input Program Analyzer Output

26 Instructor s discussion notes Input issues Input Program Analyzer Output Input is a program, but What language is the program written in? imperative vs. functional vs. object-oriented? maybe even declarative? what pointer model does the language use? reflection, exceptions, continuations? type system trusted or not? one often analyzes an intermediate language... how does one design such a language?

27 Instructor s discussion notes Input issues Input Program Analyzer Output How much of the program do we see? all? one file at a time? one library at a time? reflection Any additional inputs? any human help? profile info?

28 Instructor s discussion notes Analysis issues Input Program Analyzer Output Analysis/compilation model Separate compilation/analysis quick, but no opportunities for interprocedural analysis Link-time allows interprocedural and whole program analysis but what about shared precompiled libraries? and what about compile-time? Run-time best optimization/analysis potential (can even use run-time state as additional information) can handle run-time extensions to the program but severe pressure to limit compilation time Selective run-time compilation choose what part of compilation to delay until run-time can balance compile-time/benefit tradeoffs

29 Instructor s discussion notes Analysis issues Input Program Analyzer Output Does running-time matter? for use in IDE? or in overnight compile?

30 Instructor s discussion notes Output issues Input Program Analyzer Output Form of output varies widely, depending on analysis alias information constantness information loop terminates/does not terminate Correctness of analysis results depends on what the results are used for are we attempting to design algorithms for solving undecidable problems? notion of approximation statistical output

31 Program Transformation Issues (discuss) Input Program Transformer Output

32 Instructor s discussion notes Input issues Input Program Transformer Output A program, and Program analysis results Profile info? Environment: # of CPUs, # of cores/cpu, cache size, etc. Anything else?

33 Instructor s discussion notes Transformation issues Input Program Transformer Output What is profitable? What order to perform transformations? What happens to the program representation? What happens to the computed information? For example alias information? Need to recompute?

34 Instructor s discussion notes Output issues Input Program Transformer Output Output in same IL as input? Should the output program behave the same way as the input program?